Yesterday AMD held press event for its upcoming ZEN cpu to replace the Buldozer family.
Though AMD didn’t show official numbers, the testing was done on ZEN 8 core 16 thread verses Intel Brodwel 8 Core 16 thread (both at 3Ghz)
The test. Blender. Now that is some free advertisement for Blender by AMD side. Now I just hope OpenCL implementation gets proper attention with the new developer sponsored by AMD (I hope)
<quote>During the event, AMD demonstrated an 8-core, 16-thread “Summit Ridge” desktop processor (featuring AMD’s “Zen” core) outperforming a similarly configured 8-core, 16-thread Intel “Broadwell-E” processor[SUP]1[/SUP] when running the multi-
threaded Blender rendering software with both CPUs set to the same clock speed.</quote>
As far I know, Blender Foundation decides where to assign money from sponsors, and sponsoring cannot be targeted. To be coherent AMD should pay/hire directly a developer, and get the code contribution applied on BF blender trunk.
This is the blender rendering video where they test the new Zen processor and broadwell-E.
Both have 8 cores 16 threads at 3 Ghz. The video show the Zen processor beat broadwell-E by a bit.
It looks good. Lets hope the price will be, too. Because if the price isn’t any good, then it’s just a waste of time. Intels are good chips, but they are clearly hiking the price on purpose past what it should really go for.
A clockrate of only 3Ghz seems a bit concerning. Broadwell-E 8-core products start at 3.2Ghz base clock with a Turbo up to 3.7Ghz. Then again, those are 1000$+ chips, so if the price is right, it could still be a winner.
However, for game performance, clockrate is still more important. The question is how high the 4-core Zen chip can go (and at what price).
source -
LinusTechTips: The WAN Show - AMD Zen Benchmarks Leaked! - August 12th 2016 mentioned @ 18:50 (well worth watching from the start)
Makes common sense, Intel is abusing the market & clients for the last 5-10 years. Known practice and the problem with monopoly.
For real. I skimmed some articles but haven’t seen any mention of pricing or release dates. Anyone have any trustworthy sources? The FX-83xx series were ~$300 just a few years ago, and can be got on Newegg for under $200 now. Curious to see what these will go for…
I don’t know that that qualifies as a source, though, does it? His phrasing sounds more like hopeful speculation, or theorizing about what AMD could do to regain lost ground. I don’t think I’ll be holding my breath for a ~$200 price tag That guy on the left was hard to watch.
Indeed. AMD competition will give us all a chance to get a powerful CPU rendering power for reasonable prices.
Still I’m just happy to hear Blender’s name mentioned on nearly all computer sites. That is some exposure. Wonder if Ton was involved in any part of this, though I doubt it.
Still AMD sponsoring one person to work on Viewport speedups and supplying hardware to Blender Institute is good also. I also hope, and will continue to hope for OpenCL to support same functionality as CUDA… though that will probably take a lot of time.
This is a statement by Ton Roosendal on the subject:
After the event I met with Roy Taylor, he confirmed the support they already give to Blender developer Mike Erwin (to upgrade OpenGL). Roy said AMD is committed to help us in many more ways, so I asked for one more full time Cycles coder. Deal! Support for 1 year full time developer on Cycles to finish the ‘OpenCL split kernel’ project is being processed now. I’ll be busy hiring people the coming period!
Isn’t it true though that it’s the architecture of the chip that matters more these days than the raw clockrate?
Intel’s i-series chips for instance continue to see bumps in single-core performance with only minimal increases in the clock (however slight it is these days).
If the architecture is good enough and optimized enough, I can see Zen outperforming in single-core computing despite the clock being lower.
Testing at the same clock rate does make sense, after all, significantly improving IPC (Instructions Per Clock) was one of the main goals of Zen - and the test shows that at least for Blender, they apparently caught up to Intel.
Also, I guess they simply don’t want to disclose the clock rates that Zen will be able to archieve - that it will hit 3 GHz is obvious, so that wasn’t a huge reveal yet.
Either matters. You’re talking about things like IPC (instructions per clock) versus actual clockrate. Broadly speaking, if your IPC is 10% higher, you can clock 10% lower and achieve the same performance. However, IPC is fixed in the hardware, clockrate depends to some degree on other factors such as yield. Not all CPUs in a given batch will run stable at the same clockrate, that’s why they are binned and sold as different products, even though they are essentially the same chips. Intel has had their 14nm process running for quite a while and their clockrates did go up with every revision. AMD is just now starting with 14nm CPUs.
AMD has demonstrated with this test that their Zen chips are comparable in IPC with Intel’s CPUs, by running both chips at 3Ghz. However, that doesn’t answer the question at which clockrates the chip can really run. A 4Ghz i7 core will still outperform a 3Ghz Zen core by about 30% (ignoring memory here). This could translate into 30% more FPS in a CPU-bound game, for instance. The Intel chip they have tested against runs at least at 3.2Ghz, so why not clock the Zen chip at 3.2Ghz as well? The answer might be, 3.2Ghz isn’t fully stable right now (with 8 cores).
Now, for the quadcore to be truly competitive with Intel in terms of performance, it needs to run at 4Ghz+. Otherwise, it needs to be priced very aggressively and aggressive pricing isn’t bringing great profits.
If the architecture is good enough and optimized enough, I can see Zen outperforming in single-core computing despite the clock being lower.
Everything known so far points to it being about on par, which is already great feat considering the difference in R&D resources between AMD and Intel.
That guy on the left was hard to watch.